Altered Cingulostriatal Couplingin Obsessive–Compulsive Disorder

Jan Carl Beucke,1,2,* Christian Kaufmann,1,* Clas Linnman,3 Rosa Gruetzmann,1 Tanja Endrass,1

Thilo Deckersbach,2 Darin D. Dougherty,2 and Norbert Kathmann1

Abstract

Neurobiological models of obsessive–compulsive disorder (OCD) assume abnormalities in corticostriatal net-works involving cingulate and orbitofrontal cortices, but the connectivity within these systems is rarely addressedin experimental imaging studies in this patient group. Using an established monetary reinforcement paradigmknown to involve the cingulate cortex and the ventral striatum, the present study sought to test for altered cortico-striatal coupling in OCD patients anticipating potential punishment. The anterior midcingulate cortex (aMCC), aregion integrating negative emotion, pain, and cognitive control, was chosen as a seed region due to its particularrelevance in OCD, representing the neurosurgical target for cingulotomy, and showing increased responses to er-rors in OCD patients. Results from psychophysiological interaction analyses revealed that significantly altered,inverse coupling occurs between the aMCC and the ventral striatum when OCD patients anticipate potential pun-ishment. This abnormality links the two major contemporary neurosurgical OCD target sites, and provides directexperimental evidence of altered corticostriatal connectivity in OCD. Noteworthy, an abnormal aMCC couplingwith cortical areas outside of traditional corticostriatal circuitry was identified besides the alteration in the cingu-lostriatal pathway. In conclusion, these findings support the importance of applying connectivity methods tostudy corticostriatal networks in OCD, and favor the application of effective connectivity methods to study cor-ticostriatal abnormalities in OCD patients performing tasks that involve symptom provocation and reinforcementlearning.

Key words: anterior midcingulate cortex; loop models; obsessive–compulsive disorder; psychophysiological inter-action analysis; punishment; ventral striatum

Introduction

Neurobiological models of psychiatric disorders oftenassume abnormalities in brain loops or circuits (Kopell

and Greenberg, 2008). For example, dysfunctional cortico-basal ganglia-thalamo-cortical (CBTGC) circuits (Alexanderet al., 1986; Maia and Frank, 2011) are thought to be associ-ated with obsessive–compulsive disorder (OCD) (Del Casaleet al., 2011; Graybiel and Rauch, 2000; Kopell and Greenberg,2008; Saxena et al., 1998; Whiteside et al., 2004), a psychiatriccondition characterized by obsessive thoughts that lead tocompulsive behavior (Salkovskis, 1985). While neuroimagingstudies in OCD patients have repeatedly demonstrated al-tered regional activity in CBTGC areas during rest (Baxteret al., 1987; Swedo et al., 1989; Whiteside et al., 2004), symp-tom provocation (Breiter et al., 1996; Rauch et al., 1994; Simon

et al., 2010) and reinforcement learning (Chamberlain et al.,2008; Remijnse et al., 2006), a tendency to misinterpret univar-iate general linear model (GLM) analyses to reflect alteredCBTGC connectivity becomes evident when reviewing the lit-erature. In particular, this notion is reflected by a recent re-view on imaging findings in OCD that explicitly aims toanalyze reports of dysfunctional connectivity in CBTGC cir-cuitry in OCD, but exclusively lists studies showing abnormalregional activity of CBTGC areas in this patient group (DelCasale et al., 2011). Considering the fact that univariateGLM analyses do neither address nor inform about the inter-actions among coactivated areas (Kosslyn, 1999; Stephan,2004), the present study sought to test for altered connectivitybetween CBTGC areas based on assumptions inferred fromthe previous regional activity patterns (Del Casale et al.,2011; Graybiel and Rauch, 2000; Kopell and Greenberg,

1Department of Psychology, Humboldt-Universitat zu Berlin, Berlin, Germany.2Division of Neurotherapeutics, Massachusetts General Hospital/Harvard Medical School, Boston, Massachusetts.3Center for Pain and the Brain, Boston Children’s Hospital/Harvard Medical School, Boston, Massachusetts.*Denotes equal contribution to the article as first author.

BRAIN CONNECTIVITYVolume 2, Number 4, 2012ª Mary Ann Liebert, Inc.DOI: 10.1089/brain.2012.0078

191

2008; Saxena et al., 1998) using psychophysiological interac-tion (PPI) analysis (Friston et al., 1997).

PPI analyses compute whole-brain connectivity maps for agiven seed region (Stephan, 2004) and identify areas in whichthe connectivity with the seed significantly changes due to anexperimental factor, that is, the psychological context (Fristonet al., 1997). In healthy individuals, PPIs have been applied tostudy amygdala–prefrontal interactions in the context of fear(Banks et al., 2007; Das et al., 2005; Williams et al., 2006), andperiaqueductal gray connectivity during pain (Linnman et al.,2012) and defensive fear (Mobbs et al., 2009). Despite the sim-plicity of PPI, testing for task-specific connectivity changes ofone single region in the absence of particular model assump-tions, its applications in the context of group comparisons in-volving neuropsychiatric populations have been relativelyrare. However, PPI has been used previously to investigatealtered cingulate connectivity in post-traumatic stress disor-der (Lanius et al., 2004), amygdala–prefrontal coupling in so-cial anxiety disorder (Guyer et al., 2008), the connectivity ofmotor areas in traumatic brain injury (Kasahara et al.,2010), and self-other source monitoring in schizophrenia(Wang et al., 2011).

In OCD, the anterior midcingulate cortex (aMCC) (Vogt,2005), a brain region involved in multiple functions, such aspain, fear avoidance (Vogt, 2005), error avoidance (Magnoet al., 2006), and reinforcement-guided decision-making(Rushworth and Behrens, 2008) is of particular relevance, rep-resenting the neurosurgical target for cingulotomy (Dough-erty et al., 2002). This procedure is known to reduce striatalvolume (Rauch et al., 2000). Further, the aMCC harbors therostral cingulate zone (Shackman et al., 2011), that is, thesource of the error-related negativity (Debener et al., 2005),an electrophysiological correlate of performance monitoring,which has consistently been shown to be enhanced in OCDpatients (Endrass et al., 2008, 2010; Riesel et al., 2011). In ad-dition to electrophysiological data, neuroimaging studies ofOCD patients have also linked the aMCC with overactiveerror processing (Fitzgerald et al., 2010). Considering its ana-tomical location and connections, the aMCC has been sug-gested to represent a hub region that is ideally suited tointegrate information and to refine behaviors motivated bynegative feedback (e.g., pain or punishment) via its directconnectivity with limbic, motor, and dopaminergic midbrainareas (Shackman et al., 2011). In particular, the anatomicalconnection of the aMCC with the ventral striatum, includingthe nucleus accumbens [the cingulostriatal pathway(Kunishio and Haber, 1994)], has been linked with the coordi-nation of aversively motivated instrumental behaviors(Shackman et al., 2011).

The present study sought to ask whether altered couplingoccurs between the aMCC and other CBTGC circuit compo-nents, and the ventral striatum in particular, when OCD pa-tients experience contexts that involve errors and negativefeedback. Therefore, we tested for altered functional couplingbetween the aMCC and the ventral striatum in OCD patientsin an experimental context that involves both regions of inter-est, applying the monetary incentive delay (MID) task (Knut-son et al., 2000, 2001) in conjunction with PPI analysis.Noteworthy, a previous univariate GLM analysis of this data-set had indicated altered activation in medial prefrontal andcingulate cortices in OCD patients, but no group differenceswith respect to the regional activity of the striatum (Kauf-

mann et al., in revision). As a seed region for PPI, the presentstudy used an aMCC region located in the cingulotomy targetarea (Rauch et al., 2000), approaching areas of the aMCCwhite matter abnormalities in OCD (Szeszko et al., 2005).Based on CBTGC circuit models in OCD (Kopell and Green-berg, 2008), we predicted altered coupling in the cingulostria-tal pathway in OCD patients during anticipation of potentialpunishment, a condition that is relevant to both OCD symp-tomatology due to potentially negative consequences in thecase of erroneous performance (Alonso et al., 2008; Salkov-skis, 1985) and aMCC function (Shackman et al., 2011).

Materials and Methods

Participants

Nineteen adult outpatients with a DSM-IV diagnosis ofOCD and 19 healthy controls matched for age, gender, intel-ligence, and handedness were recruited from the OCD outpa-tient clinic at the Humboldt-Universitat zu Berlin. Patientswere interviewed by a licensed clinical psychologist and diag-nosed using the German version of the Structured ClinicalInterview for DSM-IV (First et al., 1996). Ten patients had cor-morbid diagnoses (affective disorder, n = 7; phobic disorders,n = 3; impulse control disorder, n = 3). Three out of nineteenpatients of the analysis sample were receiving medicationsto treat OCD at the time of the functional magnetic resonanceimaging (fMRI) examination (one patient took clomipramine10 mg/day, the second a combination of paroxetine 30 mg/day and clomipramine 75 mg/day, respectively, and thethird took venlafaxine 75 mg/day). No patient received ben-zodiazepines within 4 weeks before the scanning session.Exclusion criteria for patients and healthy controls includedcardiac pacemakers or other metallic implants or artifacts,and pregnancy. Patients had no significant neurological ill-ness, no previous psychosurgery, no current or past substanceabuse or dependence, dementia, delirium, schizophrenia, delu-sional disorder, or other psychotic disorder. Healthy controlshad neither DSM-IV diagnosis (as indicated by a negative pa-tient report in a SCID screening questionnaire) nor a first-de-gree relative with a reported history of psychiatric disorders.All participants provided written informed consent accordingto the institutional guidelines before enrollment. t-Tests for in-dependent samples were used to compare groups regardingage, IQ, handedness, earnings in the MID task, and STAI testscores. t Values, means and standard deviations of question-naire data are available in Table 1. The study was approvedby the local ethics committee.

MID task

The MID task (Knutson et al., 2001) was adapted in thepresent study (Fig. 1). There were seven different types of tri-als: three trial-types with the possibility of winning money ona correct (i.e., timely) button press (potential reward trials),three trial-types with the possibility to avoid losing moneyon a correct button press (potential punishment trials), andthe remaining trial-type in the absence of any monetaryconsequences (neutral trials). At the beginning of each trial,one of seven different cues was shown to indicate trial type.Participants were asked to press a button as soon as the tar-get stimulus (gray colored square) appeared. Depending onthe performance (i.e., timely motor response), participants

192 BEUCKE ET AL.

received feedback about winning or losing money. Each runconsisted of 72 trials, that is, 27 indicating potential reward,27 indicating potential punishment, and 18 neutral cueswith each trial lasting for 11.6 sec on average (see Fig. 1 fortiming details). The interstimulus interval (from target offsetto cue onset) ranged from 6570 to 7370 msec. The experimentlasted for 14 min (depending on the subject’s response time).Task difficulty was continuously adapted to achieve a 66%success level in each subject across a task run, which was de-rived from individually tailored reaction times from the train-ing session, and by adapting the response deadline as a resultof the averaged reaction times in previous trials and the cor-rectness of the immediately preceding trial during the test runs.

Clinical scales

Severity of OCD symptoms was evaluated using the Ger-man versions of the Yale-Brown Obsessive–CompulsiveScale (Goodman et al., 1989) and the Obsessive–CompulsiveInventory Revised (Foa et al., 2002). In addition, the Stateand Trait Anxiety Inventory-Revised (Spielberger et al.,1970) and the Beck Depression Inventory (Beck et al., 1961)were administered before scanning. The Edinburgh Handed-ness Inventory (Oldfield, 1971) was used to classify handed-ness. For matching purposes, all participants completeda German vocabulary test (Schmidt and Metzler, 1992) as abrief measure of verbal intelligence.

Table 1. Demographic and Psychometric Data and Total Earnings in the Monetary Incentive Delay Taskof Obsessive–Compulsive Disorder Patients and Matched Healthy Controls

N = 19 vs. 19 OCD (M – SD) Control t (p value)

Sex [female (male)] 11 (8) 11 (8)Age 34.8 (11.0) 34.9 (11.8) 0.03 (0.98)Intelligence (verbal) 104 (10) 107 (12) !1.07 (0.29)Handedness 77 (55) 67 (47) 0.65 (0.52)STAI-X1 (state) 54 (13) 49 (6) 1.61 (0.12)STAI-X2 (trait) 61 (13) 50 (6) 3.32 (0.002)Earnings in e 21.80 (7.80) 22.00 (6.40) !0.07 (0.95)Y-BOCS 20.7 (7.9)OCI-R total score 23 (13)

OCI-R checking subscore 4 (4)OCI-R hoarding subscore 3 (4)OCI-R neutralization subscore 2 (3)OCI-R obsession subscore 6 (4)OCI-R symmetry subscore 4 (4)OCI-R washing subscore 4 (4)

BDI 17 (11)Medication (N) 3Comorbidity (N) 10

Comparisons are based on two-sample t tests. More information regarding symptom dimensions is available in Supplementary Tables S1aand b. (Supplementary Data are available online at www.liebertpub.com/brain).

OCD, obsessive–compulsive disorder; SD, standard deviation; STAI, State and Trait Anxiety Inventory; Y-BOCS, Yale-Brown Obsessive–Compulsive Scale; OCI-R, Obsessive–Compulsive Inventory Revised; BDI, Beck Depression Inventory.

FIG. 1. Monetary incentive delaytask. (a) The diagram shows thedifferent trial types and (b) thecourse of one trial. In this example,the subject did not respond in time,and therefore was punished with areduction of the total winning sumby 3 Euro (i.e., winnings werereduced from e 12.60 to e 9.60).Participants started with a credit of5 Euro; the possible maximum sumof winnings was e 38.30.

ALTERED CINGULOSTRIATAL COUPLING IN OCD 193

fMRI acquisition and statistical analysis

Stimuli were generated using Presentation (Neurobehavio-ral Systems) and were projected by means of a mirror systemattached to the head coil. To reduce head motion, the subjects’head was immobilized by a vacuum head cushion. Earplugswere used to attenuate background noise. Headphones wereused to communicate with subjects. Before functional runs,176 anatomical Modified Driven Equilibrium Fourier Trans-form (MDEFT) slices were acquired (spatial resolution1 · 1 · 1 mm, repetition time (TR) = 12.24 msec, echo time(TE) = 3.56 msec, flip angle = 23!, 256 · 224 matrix) duringthe training session of the MID task on a 1.5 T Siemens Sonatascanner. A total of 450 volumes (T2*-weighted single-shotgradient echo planar imaging sequence) were acquired ineach of the two test runs using the following parameters:TR = 1870 msec, TE = 40 msec, 33 consecutive slices,3 · 3 · 3.5-mm voxel, flip angle = 90!, field of view (FOV) = 192mm, 64 · 64 matrix.

fMRI data analysis was performed using SPM5 (Version1782, Statistical Parametric Mapping, www.fil.ion.ucl.ac.uk).First, the original data files were converted from Dicom toNifti file format. The first three volumes of each functionaltime series were discarded to avoid nonsteady state effectscaused by T1 saturation. After slice time correction, we real-igned all volumes to the first volume to correct for between-scan movements and to remove signals correlated withhead motion. Motion correction estimation revealed that nosubject showed more than 2-mm head movement and morethan one degree of rotation during one run. The anatomicaldata set was coregistered with the mean T2* image and T1-weighted images were segmented into gray matter, whitematter, and cerebrospinal fluid. The gray matter of the core-gistered structural image was spatially normalized to thestandard template provided by the Montreal NeurologicalInstitute (MNI) template using an automated spatial transfor-mation (12-parameter affine transformation followed by non-linear iterations using 7 · 8 · 7 basis functions). The resultingtransformation matrix was subsequently applied to the T2*data, and a resampling to a resolution of 3 · 3 · 3-mm voxelsize was performed. Finally, the normalized images weresmoothed with a Gaussian kernel (full width at half maxi-mum) of 8 mm to create a locally weighted average of the sur-rounding voxels. The GLM was applied by modeling onsetsand durations of anticipation and feedback of the three differ-ent conditions as regressors (punishment, reward, and neu-tral). These regressors were convolved with a canonicalhemodynamic response function as implemented in SPM5.Thus, the three different cue conditions (reward, punishment,neutral) were modeled, and noteworthy, the different degreesof reward and punishment were linearly weighted within aparametric design.

PPI analysis

PPI analyses test whether one region, the seed time series,changes its coupling with other regions due to a specific ex-perimental manipulation (Friston et al., 1997; Gitelmanet al., 2003). To perform PPI analysis, seed voxels were de-rived by extracting the first eigenvariate time series (volumesof interest within SPM5) from a 5-mm sphere centeredaround three aMCC regions. Volumes of interest were ad-justed for effects of interest. In the present study, seed voxels

were extracted from the aMCC gray matter regions encom-passing both the cingulotomy target region in Brodmannarea 24 (Dougherty et al., 2002; Rauch et al., 2000) and regionswhere functional anisotropy decreases have previously beendescribed in OCD (Szeszko et al., 2005). Since these coordina-tes were located in the white matter, the nearest gray matterlocus displaying individual task effects was used for seedvoxel extraction. Three control steps were conducted to en-sure validity of seed voxel extraction (see SupplementaryMethods for details; Supplementary Data are available onlineat www.liebertpub.com/brain), revealing an aMCC seedBrodmann area 24 (MNIXYZ =!6, 15, 30; Fig. 2b).

PPI analyses regress the interaction between a BOLD seedvoxel time series (here: aMCC seed) and a psychological pa-rameter (anticipation of punishment minus anticipation of re-ward) on all voxel time series in the brain (Friston et al., 1997;Gitelman et al., 2003). Therefore, the seed time series and thepsychological factor are multiplied (element-by-element mul-tiplication), resulting in the PPI interaction term regressor.Accordingly, the design matrix for PPI analyses includesthree columns: the psychological parameter (onsets of antici-pated punishment and reward), the seed voxel time series(physiological parameter), and the PPI interaction term. Mod-eling the interaction term in the first column, and using thecontrast [1 0 0] allowed to ask for voxel time series that arepredicted by the PPI interaction term regressor. Importantly,the extracted time series was deconvolved with a canonicalhemodynamic response function before creating the interac-tion term, to account for the fact that interaction of brain re-gions are not expressed on the level of hemodynamicresponses, but on a neuronal level (Gitelman et al., 2003). Inaddition, six motion correction parameters were includedinto the PPI design matrix representing effects of no interest.

PPI group analysis

For each subject, PPI effects were estimated at each voxel,and statistical parametric maps (SPMs) were calculated forthe psychological parameter contrast of interest (anticipatedpunishment minus anticipated reward). These individualSPMs were submitted to a random effects analysis, and atwo-sample t test was performed to compare PPIs betweengroups. Due to the fact that the psychological parameter rep-resents a term subtracting punishment (i.e., [1]) and reward(i.e., [!1]) conditions, the PPI regressor asks for voxel time se-ries that are predicted by the seed during punishment,whereas signaling inverse to the seed time series is predictedin the reward condition. Therefore, the PPI regressor identi-fies voxels that show a difference in regression slopes whencomparing regressions of the seed time series between thetwo conditions, indicating that a change in task conditionshas an effect on the connectivity among the seed and the iden-tified region. This difference in regression slopes is subse-quently compared between two groups, and accordingly,regions showing significant PPI group effects (two-samplet test) reveal significant group differences with respect tothe differences in regression slopes. For this reason, explana-tory PPI analyses are needed for reasonable interpretation ofPPI group effects, allowing one to attribute the PPI group ef-fects to one group, and ideally, one condition driving it. Tocharacterize PPI differences between groups in detail, effectsare provided using an uncorrected threshold of p < 0.001 with

194 BEUCKE ET AL.

an extent of 5 voxels (Tables 2 and 3). In accordance withprevious PPI studies (Das et al., 2005; Mobbs et al., 2009),we performed small volume corrections for a priori regionsof interest, using a threshold of p < 0.05 (false discovery ratecorrected for multiple comparisons). These were appliedusing one combined, bilateral mask of the following regionsin all analyses: caudate, putamen, pallidum (as implementedin the Automatic Anatomic Labeling tool integrated in theWFU PickAtlas (www.fmri.wfubmc.edu/cms/software/Pickatlas).

Explanatory PPI analyses

Within one group, positive beta values indicate a positiverelationship between the seed and an identified region inone condition relative to the other (positive PPI), whereasnegative beta values index a negative relationship in one con-dition relative to the other (negative PPI). Accordingly, com-paring effects in the OCD group to the control group revealsmore positive (OCD > Control) or more negative (Control >OCD) beta values in the OCD group. However, the interpre-tation of PPI group effects is hindered by the fact that differ-ence values of two groups are subtracted from each other. Asfor univariate GLM group comparisons, a second level groupeffect using individual first level contrast images (already in-

cluding a subtraction of two conditions) as an input cannot bedirectly attributed to a single group or condition driving it. Toovercome this problem, two additional explanatory PPI ana-lyses were conducted. Thus, regressions of the original aMCCseed time series for both conditions (punishment, reward)were calculated separately, comparing one condition at atime to the implicit baseline, instead of directly subtractingthem from one another as in the original PPI analysis. Thisresulted in two new first level PPI contrasts in each subject(anticipated punishment > implicit baseline, anticipated re-ward > implicit baseline). From a technical perspective, theonly difference between the original PPI first level analysisand the explanatory PPI first level analysis was that task pa-rameters (punishment or reward) were removed from thepsychological parameter (i.e., [1 0] and [0 1] instead of [1!1]) in the original design matrix) one at a time before calcu-lating the PPI interaction term. Thus, two new interactionterm regressors were built and accordingly, two new firstlevel models were estimated, resulting in two separateSPMs (i.e., anticipation of punishment > implicit baseline, an-ticipation of potential reward > implicit baseline) for each par-ticipant. Subsequently, these new individual contrast imageswere fed into random effects analyses. This allowed to per-form group comparisons (two-sample t tests as before) withrespect to the potential punishment condition. The attempt

FIG. 2. PPI group effects in the bilateral ventral striatum, displayed coronally and axially (a); aMCC seed region in Brodmanarea 24 (in red, MNIXYZ =!6, 15, 30), and univariate GLM task effects of the combined sample (in yellow, n = 38, anticipation ofreward > neutral anticipation, p < 0.05, false discovery rate corrected) (b); and boxplots displaying PPI beta estimates indexingcingulostriatal coupling for potential punishment minus potential reward in both groups (c) revealing positive beta values incontrols and negative beta values in OCD patients (Controls > OCD). aMCC, anterior midcingulate cortex; VS, ventromedialstriatum; PPI, psychophysiological interaction; MNI, Montreal Neurological Institute; OCD, obsessive–compulsive disorder;GLM, general linear model. M, mean; SEM, standard error mean; HC, healthy controls.

ALTERED CINGULOSTRIATAL COUPLING IN OCD 195

Ta

ble

2.R

egio

ns

Dem

on

str

ati

ng

Sig

nifi

ca

nt

Psy

ch

oph

ysio

log

ica

lIn

ter

ac

tio

nG

ro

up

Eff

ects

wit

hth

eA

nte

rio

rM

idc

ing

ula

teC

or

tex

Seed

Reg

ion

Du

rin

gA

nti

cip

ati

on

of

Po

ten

tia

lP

un

ish

men

t(M

inu

sP

ote

nti

al

Rew

ar

d)

Con

trol

sP

atie

nts

Dif

fere

nce

a

An

atom

ybS

tati

stic

cA

nat

omy

Sta

tist

icA

nat

omy

Sta

tist

ic

PP

Ief

fect

regi

onD

irec

tion

dx,

y,z

zS

core

CS

BA

Dir

ecti

onx,

y,z

zS

core

CS

BA

Dir

ecti

onx,

y,z

zS

core

CS

BA

Infe

rio

rp

arie

tal

lob

ule

Neg

ativ

e!

42,!

36,

364.

7229

40H

C>

OC

D!

45,!

39,

394.

0829

40P

recu

neu

sN

egat

ive

!27

,!

48,

393.

885

31N

egat

ive

!27

,!

48,

543.

6610

7P

ost

cen

tral

gy

rus

Neg

ativ

e!

48,!

24,

543.

8057

40H

C>

OC

D45

,!

27,

603.

907

40H

C>

OC

D!

27,!

30,

573.

855

40H

C>

OC

D!

48,!

24,

513.

517

40P

ost

erio

rci

ng

ula

teN

egat

ive

9,!

54,

213.

508

23V

entr

alst

riat

um

Neg

ativ

e!

15,

15,!

124.

4111

2H

C>

OC

D!

15,

15,!

93.

95e

30N

egat

ive

12,

9,!

33.

8736

HC

>O

CD

15,

9,!

93.

62e

14P

uta

men

Neg

ativ

e!

30,!

18,!

63.

8610

Cau

dat

eh

ead

Neg

ativ

e!

9,9,

93.

456

Cla

ust

rum

Neg

ativ

e30

,21

,!

93.

357

Med

ial

fro

nta

lg

yru

sN

egat

ive

0,!

12,

603.

7338

6H

C>

OC

D6,!

6,54

3.48

56

Neg

ativ

e6,!

12,

693.

7115

6N

egat

ive

0,3,

663.

325

6A

nte

rio

rci

ng

ula

teN

egat

ive

!6,

27,!

93.

7624

9M

idd

lefr

on

tal

gy

rus

Neg

ativ

e!

39,

57,

03.

685

10P

rece

ntr

alg

yru

sN

egat

ive

30,!

15,

663.

8832

4M

idd

lete

mp

ora

lg

yru

sN

egat

ive

54,!

66,

184.

0321

39O

CD

>H

C54

,!

63,

183.

9117

39Su

per

ior

tem

po

ral

gy

rus

Neg

ativ

e!

42,!

54,

214.

0420

22C

un

eus

Neg

ativ

e24

,!

75,

213.

5510

18N

egat

ive

0,!

87,

243.

398

18C

ereb

ellu

m(c

ulm

en)

Neg

ativ

e21

,!

51,!

154.

6112

5H

C>

OC

D0,!

51,!

93.

9215

Neg

ativ

e!

6,!

48,!

93.

367

Cer

ebel

lum

(dec

liv

e)N

egat

ive

6,!

78,!

154.

0322

Clu

ster

size

sin

dic

ate

the

nu

mb

ero

fco

nti

gu

ou

sv

oxe

lsfo

rea

chcl

ust

er.

aB

etw

een

-gro

up

effe

cts

thre

sho

lded

atp

<0.

001

(un

corr

ecte

d;

min

imu

mcl

ust

erex

ten

t,fi

ve

con

tig

uo

us

vo

xels

).bP

PI

effe

ctco

ord

inat

es(x

,y,

z)ar

eg

iven

inM

NI

spac

e.c M

agn

itu

de

stat

isti

csco

rres

po

nd

toa

min

imu

m(w

hol

e-b

rain

)u

nco

rrec

ted

thre

sho

ldo

fp

<0.

001

wit

ha

min

imu

mcl

ust

erex

ten

to

ffi

ve

con

tig

uo

us

vo

xels

.dD

irec

tio

no

fth

eP

PI:

neg

ativ

ein

dic

atin

ga

mo

ren

egat

ive

rela

tio

nsh

ipb

etw

een

seed

and

PP

Ief

fect

reg

ion

du

rin

gp

un

ish

men

tre

lati

ve

toth

ere

war

dco

nd

itio

n.

e Sig

nifi

can

tbet

wee

n-g

rou

pm

ain

effe

ctd

iffe

ren

ces

ina

pri

ori

reg

ion

so

fin

tere

st(o

ne

com

bin

edm

ask

ofb

ilat

eral

cau

dat

e,p

uta

men

,an

dp

alli

du

mm

ask

sg

ener

ated

fro

mth

eA

AL

too

loft

he

Wak

eF

ore

stU

niv

ersi

tyW

FU

Pic

kA

tlas

[htt

p:/

/fm

ri.w

fub

mc.

edu

/so

ftw

are/

Pic

kA

tlas

])u

sin

ga

corr

ecte

dth

resh

old

of

p FD

R<

0.05

.B

A,B

rod

man

nar

ea;H

C,h

ealt

hy

con

tro

ls;C

S,cl

ust

ersi

ze,P

PI,

psy

cho

ph

ysi

olo

gic

alin

tera

ctio

n;M

NI,

Mo

ntr

ealN

euro

log

ical

Inst

itu

te;A

AL

,Au

tom

atic

An

ato

mic

Lab

elin

g;F

DR

,fal

sed

isco

ver

yra

te.

196

Ta

ble

3.E

xpla

na

tor

yP

syc

ho

phys

iolo

gic

al

Inte

ra

cti

on

Eff

ects

wit

hth

eA

nte

rio

rM

idc

ing

ula

teC

or

tex

Seed

Reg

ion

Du

rin

gA

nti

cip

ati

on

of

Po

ten

tia

lP

un

ish

men

t(V

ersu

sIm

plic

itB

ase

lin

e)

Con

trol

sP

atie

nts

Dif

fere

nce

a

An

atom

ybS

tati

stic

cA

nat

omy

Sta

tist

icA

nat

omy

Sta

tist

ic

PP

Ief

fect

regi

onD

irec

tion

dx,

y,z

zS

core

CS

BA

Dir

ecti

onx,

y,z

zS

core

CS

BA

Dir

ecti

onx,

y,z

zS

core

CS

BA

Occ

ipit

alg

yru

s(l

ing

ual

)N

egat

ive

27,!

57,!

124.

4114

319

HC

>O

CD

18,!

63,!

63.

9522

19N

egat

ive

27,!

63,

63.

615

19C

un

eus

Neg

ativ

e27

,!

84,

303.

487

18N

egat

ive

!3,!

81,

393.

8215

19N

egat

ive

!15

,!

81,

333.

599

18O

rbit

ofr

on

tal

gy

rus

Neg

ativ

e36

,33

,!

64.

0658

47H

C>

OC

D36

,33

,!

63.

7728

47N

egat

ive

!27

,27

,!

183.

6119

47V

entr

alst

riat

um

Neg

ativ

e!

9,12

,!

33.

8292

HC

>O

CD

15,

9,!

94.

63e

60H

C>

OC

D!

9,12

,!

33.

81e

17P

ulv

inar

Neg

ativ

e!

9,!

27,

33.

5714

Insu

laN

egat

ive

!39

,!

15,

153.

489

13M

edia

lfr

on

tal

gy

rus

Neg

ativ

e0,!

15,

603.

6615

6N

egat

ive

9,51

,6

3.39

119

Pre

cen

tral

gy

rus

Neg

ativ

e!

15,!

12,

693.

697

6P

arac

entr

allo

bu

leN

egat

ive

9,!

24,

663.

778

6Su

pra

mar

gin

alg

yru

sN

egat

ive!

45,!

72,!

183.

275

40H

C>

OC

D!

45,!

39,

393.

556

40M

idd

lete

mp

ora

lg

yru

sN

egat

ive

66,!

48,

03.

789

21Su

per

ior

tem

po

ral

gy

rus

Neg

ativ

e!

42,!

30,

63.

605

41R

edn

ucl

eus

Neg

ativ

e9,!

24,!

153.

728

HC

>O

CD

6,!

21,!

123.

517

Cer

ebel

lum

(cu

lmen

)H

C>

OC

D!

15,!

45,!

213.

8117

Cer

ebel

lum

(dec

liv

e)N

egat

ive

9,!

78,!

183.

6525

HC

>O

CD

0,54

,!

123.

466

Neg

ativ

e!

15,!

57,!

123.

649

Neg

ativ

e45

,!

72,!

183.

489

Clu

ster

size

sin

dic

ate

the

nu

mb

ero

fco

nti

gu

ou

sv

oxe

lsfo

rea

chcl

ust

er.

aB

etw

een

-gro

up

effe

cts

thre

sho

lded

atp

<0.

001

(un

corr

ecte

d;

min

imu

mcl

ust

erex

ten

t,fi

ve

con

tig

uo

us

vo

xels

).bP

PI

effe

ctco

ord

inat

es(x

,y,

z)ar

eg

iven

inM

NI

spac

e.c M

agn

itu

de

stat

isti

csco

rres

po

nd

toa

min

imu

m(w

ho

le-b

rain

)u

nco

rrec

ted

thre

sho

ldo

fp

<0.

001

wit

ha

min

imu

mcl

ust

erex

ten

to

ffi

ve

con

tig

uo

us

vo

xels

.dD

irec

tio

no

fth

eP

PI:

neg

ativ

ein

dic

atin

ga

neg

ativ

ein

tera

ctio

nb

etw

een

seed

and

PP

Ief

fect

reg

ion

du

rin

gp

ote

nti

alp

un

ish

men

t.e Si

gn

ifica

ntb

etw

een

-gro

up

mai

nef

fect

dif

fere

nce

sin

ap

rio

rire

gio

ns

of

inte

rest

(on

eco

mb

ined

mas

ko

fbil

ater

alca

ud

ate,

pu

tam

en,a

nd

pal

lid

um

mas

ks

gen

erat

edfr

om

the

AA

Lto

olo

fth

eW

ake

Fo

rest

Un

iver

sity

WF

UP

ick

Atl

as[h

ttp

://

fmri

.wfu

bm

c.ed

u/

soft

war

e/P

ick

Atl

as])

usi

ng

aco

rrec

ted

thre

sho

ldo

fp F

DR

<0.

05.

to decompose PPI effects might be complicated by the factthat PPI group analyses compare regression slopes betweenconditions, whereas explanatory PPI analyses compare the re-gression slope of single conditions to an implicit baseline(grand mean). Therefore, we validated the explanatory PPIprocedure by correlating PPI beta estimates from the originalPPI analysis with difference (i.e., subsequently subtracted)values of beta estimates from explanatory PPI analyses, as-suming that a high correlation between these two variableswould indicate successful decomposition of the PPI group ef-fect. To validate successful decomposition of the PPI group ef-fects using explanatory PPI analysis, the combined sample(n = 38) was used to calculate Pearson correlation coefficientsbetween the PPI beta estimates of the left (MNIXYZ =!15, 18,!9) and right (MNIXYZ = 15, 9,!6) ventral striatal peak voxelsof significant PPI group effects and difference values of thePPI beta estimates of the same exact coordinate from the ex-planatory PPI analyses (PPI beta estimates potential loss of re-ward > implicit baseline minus PPI beta estimates potentialreward > implicit baseline), revealing a highly significant cor-relation (left ventral striatum: r = 0.97, R2 = 0.94; right ventralstriatum: r = 0.84, R2 = 0.71), assuring accuracy and validityof the explanatory PPI results.

Results

PPI results

PPI group analysis revealed significant group differences(two-sample t-test, Controls > OCD, potential punishmentminus potential reward) for the ventral striatum bilaterally(Fig. 2a), that is, ventromedial left caudate head and rightputamen. Additionally, PPI group differences were observedbetween the aMCC seed and inferior, postcentral, and parie-tal areas, and further medial prefrontal and cerebellar areas inthe patient group compared with healthy controls (Controls >OCD; Table 2). These PPI effects for Controls > OCD indicatethat the relationship between the aMCC and PPI effect re-gions is more positive in the reward compared with the pun-ishment condition in Controls in relation to patients, or viceversa, a more negative relationship between the aMCC andPPI effect regions in the punishment compared with the re-ward condition in patients in relation to controls. Therefore,reasonable interpretation of PPI group differences requiresconsideration of within-group (main) effects (Table 2) andadditional explanatory PPI analyses (see below). PPI groupdifferences for OCD > Controls were observed for a regionin the middle temporal gyrus. Analyses considering medica-tion status and potential partial white matter signalextraction as potential confounds are available as Supple-mentary Materials (Supplementary Tables S2–S5 and Sup-plementary Figs. S1 and S2). There was no significantcorrelation between PPI effects and OCD symptom severityor duration of OCD in years (see Supplementary Methodsand Results).

Explanatory PPI analyses

Comparisons between groups regarding aMCC connectiv-ity in the potential punishment condition (relative to implicitbaseline, derived using explanatory PPI analyses revealedsignificant group differences (two-sample t-tests) for cingu-lostriatal coupling (Controls > OCD), that is, right ventralputamen extending into nucleus accumbens (Fig. 3a and

Table 3) and left ventral caudate head (Fig. 3a). Besidesgroup differences in basal ganglia regions, inverse couplingbetween the aMCC and orbitofrontal, inferior parietal, lingualand cerebellar areas and the red nucleus was evident. All theabove effects were observed for Controls > OCD, indicatingnegative beta values in patients compared with controls. Apost hoc search for group differences in aMCC connectivitywith the midbrain during potential punishment was con-ducted, since dopaminergic centers are critical in modulatingprediction error responses in the aMCC and the ventral stria-tum (Rushworth and Behrens, 2008). This analysis revealedthat negative coupling effects similar to the ones observedin the cingulostriatal pathway were also present betweenthe aMCC and the substantia nigra, given that the red nucleuscluster extended into this region when using a liberal thresh-old of p < 0.005, uncorrected (Fig. 4).

Discussion

This study tested for aMCC connectivity abnormalities inOCD patients anticipating potential punishment. PPI analy-ses revealed significantly altered, inverse coupling of theaMCC with the bilateral ventral striatum and inferior parietalregions known to exhibit direct anatomical connections withthe aMCC (Kunishio and Haber, 1994; Vogt, 2005; Vogt andPandya, 1987). More precisely, cingulostriatal coupling wasidentified to be specifically driven by an inverse relationshipof these two areas in the potential punishment condition,as revealed by explanatory PPI analyses. With respect toCBTGC model assumptions, these data extend the existing lit-erature by showing abnormal connectivity among CBTGCcomponents known to conjointly exhibit regional hyperactiv-ity during symptom provocation (Breiter et al., 1996; Rauchet al., 1994), while at the same time showing abnormal con-nectivity between the aMCC and regions showing functionalabnormalities in OCD outside traditional CBTGC circuitry(Menzies et al., 2008a). In summary, these findings suggestthat dysfunctional functional coupling in CBTGC circuitsaffects cingulostriatal pathways when OCD patients antici-pate potential punishment, while at the same time providingevidence for the view that the aMCC interacts not onlywith striatal, but rather with a more distributed corticalset of areas in OCD patients experiencing symptom-relevantcontexts.

In addition to the observation of altered coupling in thecingulostriatal pathway, it is worth noting that corticalareas showing altered inverse coupling with the aMCC dur-ing anticipation of potential punishment included the orbito-frontal cortex (OFC). This area is of particular relevance forthe pathophysiology of OCD because it is known to reveal in-creases in activity during symptom provocation (Breiter et al.,1996) and decreased activity during reversal learning in OCD(Chamberlain et al., 2008). Consistent with CBTGC circuitconsiderations involving the OFC (Kopell and Greenberg,2008), one of the few OCD studies applying connectivitymethods on fMRI data found increased resting-state func-tional connectivity between the ventral striatum and theOFC (Harrison et al., 2009). Crucially, the ventral striatal/nu-cleus accumbens coordinates used as seeds by Harrison et al.(2009) highly correspond to the areas showing abnormal con-nectivity with the aMCC in the present study. Further indica-tion for specific interplay of the aMCC, the ventral striatum,

198 BEUCKE ET AL.

and OFC in OCD is provided by the fact that all three regionshave repeatedly been shown to respond during symptom-provocation (Breiter et al., 1996; Rauch et al., 1994). In addi-tion, structural MRI analyses have previously pointed atabnormalities in the ventral striatum (Pujol et al., 2004), and

volume reductions exclusively in the bilateral striatum fol-lowing cingulotomy (Rauch et al., 2000), and a meta-analyticstudy (Rotge et al., 2009) confirmed volume reductions of theleft aMCC and bilateral OFC in OCD. Considering anatomi-cal evidence indicating that limbic loops transfer informationamong each other (Haber, 2003), these findings might suggestinteractions among CBTGC areas, namely, the aMCC, OFC,and ventral striatum in symptom-relevant contexts in OCD.

In addition to altered coupling among CBTGC regions tra-ditionally involved in the pathophysiology of OCD, alteredaMCC connectivity with parietal and lingual occipital areas,thus regions outside of CBTGC circuitry was evident. Theseareas partially overlap with locations of white matter reduc-tions in OCD, that is, inferior parietal (Menzies et al., 2008b;Szeszko et al., 2005) and lingual (Szeszko et al., 2005) areas(Tables 2 and 3), and also converge with meta-analytic evi-dence of functional alterations outside of traditional CBTGCcircuitry, predominantly found in parietal, medial occipital,and cerebellar areas (Menzies et al., 2008a).

From a methodological perspective, this is one of the fewstudies (Guyer et al., 2008; Kasahara et al., 2010; Laniuset al., 2004) that used PPI to compare diagnostic groups inthe neuropsychiatric field, in contrast to a majority of studiesapplying univariate GLM analyses on experimental fMRIdata. As already indicated above, it is crucial to apply connec-tivity methods, such as PPI, in the case of theoretical

FIG. 4. Results of a post hoc small volume search ( p < 0.005,uncorrected) for the midbrain revealing differences (Con-trols > OCD) with respect to coupling between aMCC seedand SN (MNIXYZ = 12,!15, !15, t = 3.13) and the red nucleusdue to inverse coupling between these two regions in OCDpatients anticipating potential punishment. SN, substantianigra.

FIG. 3. Explanatory PPI effects. Regressions of the aMCC seed time series during potential punishment (minus implicit base-line) compared between both groups (a), revealing significant ventral striatal effects, and within OCD patients alone (b), re-vealing inverse coupling of the aMCC with ventral striatal (including nucleus accumbens), orbitofrontal, and occipital(lingual) regions. Explanatory PPI effects allowed to plot beta estimates of the peak voxels of PPI group effects (i.e., left:MNIXYZ =!15, 15,!9, right: MNIXYZ = 15, 9,!9; Fig. 1 and Table 2), separately for potential punishment (vs. implicit baseline)and potential reward (vs. implicit baseline) conditions (c, d).

ALTERED CINGULOSTRIATAL COUPLING IN OCD 199

assumptions of altered networks (Stephan, 2004). More pre-cisely, univariate GLM and connectivity analyses are basedupon two fundamentally different concepts of brain function-ing, the first reflecting the idea of functional segregation ofbrain areas (specific regional responses to a task component),and the latter reflecting functional integration (specific inter-regional interaction to a task component) (Friston, 2011). Fol-lowing the notion that identification of candidate elements ofa neural system by means of GLM-based analyses requiressubsequent analysis of their functional integration to providea model for the underlying neural system (Stephan, 2004), thepresent study tested for altered corticostriatal coupling inOCD patients using a seed in the aMCC. The fact that the ob-served effects concern regions sharing direct anatomical con-nections with the aMCC (i.e., inferior parietal areas, ventralstriatum, substantia nigra) advocates for the view that PPI,despite its simplicity, is a useful approach to ask for whole-brain connectivity changes (Friston et al., 1997) of one partic-ular, extensively connected functional hub region as it is thecase for the aMCC (Shackman et al., 2011). Taken together,previous univariate GLM analyses have provided evidencefor regional functional abnormalities of the cingulate, includ-ing the aMCC, and the present data suggest that the anticipa-tion of potential punishment evokes abnormal functionalintegration among the aMCC and areas within (the ventralstriatum) and outside (lingual and inferior parietal gyri) ofCBTGC circuitry. While these findings of abnormal cingulateconnectivity might suggest that regional functional abnor-malities in occipital and parietal areas (Menzies et al.,2008a) are influenced by activity in the aMCC, this questionrequires application of effective connectivity approaches,such as dynamic causal modeling [DCM (Friston et al.,2003)], which directly address causal interregional interactionin response to task manipulations. Noteworthy, the presentPPI data suggest to consider the inclusion of areas outsideof traditional CBTGC circuitry in the DCM modeling, partic-ularly in the case that the applied task is known to involvemedial occipital and parietal areas.

From a clinical perspective, the finding of altered connec-tivity in the cingulostriatal pathway has implications for neu-rosurgical and stimulation treatment in OCD. This becomesevident considering that an aMCC area located in the cingu-lotomy target region (Dougherty et al., 2002; Rauch et al.,2000) was used as a seed, whereas the ventral striatum, themain PPI effect region, represents a primary target for deepbrain stimulation (DBS) in OCD (Denys et al., 2010; Rauchet al., 2006). Remarkably, recent studies emphasize particularefficiency of using the anterior internal capsule, that is, the an-atomical connection between the aMCC and the ventral stria-tum (Kopell and Greenberg, 2008), as a target for stimulationand surgery (Ruck et al., 2008). Thus, the present findingsmight provide potential explanations for the observationthat targeting of different neuroanatomical locations hasbeen shown to be beneficial for refractory OCD patients,since these treatments all influence locations within the cingu-lostriatal pathway. Future research on DBS stimulation inOCD might benefit from the observation of altered connectiv-ity between two different target locations, and from exami-nations of how DBS affects cingulostriatal connectivity.Similarly, studying how cingulostriatal stimulation affectsthe OFC and the above-mentioned areas outside CBTGC cir-cuitry revealing functional alterations (Menzies et al., 2008a)

could help to further specify neuroanatomical models ofOCD.

Limitations of the present findings concern the small sam-ple size, and the fact that PPI analyses involve pair-wise re-gressions of a single brain region, thus limiting conclusionsabout complex dynamic interactions encompassing multiplebrain regions. Further, PPI do not inform about causalityamong influences between brain regions, favoring the appli-cation of DCM analyses (Stephan et al., 2010) in future exam-inations of corticostriatal connectivity in OCD. Furtherlimitations of the present study concern the application offMRI as a method that does not address neurochemical mech-anisms that might underly abnormal cingulostriatal coupling.While the experimental context of this study (Knutson et al.,2000, 2001) and post hoc analyses revealing negative aMCC-substantia nigra coupling suggest dopaminergic involve-ment, it is known that direct, reciprocal glutamatergicconnections exist between the aMCC and the ventral striatum(Kopell and Greenberg, 2008), which might also contribute tothe abnormalities found here. Depletion and neurochemicaltracer studies could potentially reveal specific transmitterinterplay in the cingulostriatal pathway in relation toobsessive–compulsive behavior.

In summary, our results reveal specific aMCC connectivityabnormalities in OCD patients in an experimental contextthat challenges aMCC function. Our finding of inverse cingu-lostriatal coupling demonstrates direct evidence of a dysfunc-tional pathway within CBGTC circuits, while at the same timeidentifying altered aMCC coupling with cortical areas outsideof traditional CBTGC circuitry. Therefore, the present find-ings relate to neurobiological models of OCD. Further, thesedata are clinically relevant considering that contemporaryneurosurgical OCD treatments target components of the cin-gulostriatal pathway. Finally, the present data favor the ap-plication of effective connectivity approaches to studycorticostriatal connectivity in OCD.

Acknowledgments

Jan Beucke is supported by a Ph.D. scholarship from Evan-gelisches Studienwerk e.V. Villigst (Schwerte, Germany), andis also an ERP scholar of the German National AcademicFoundation. The authors thank Eva Kischkel, Ph.D., andRudiger Spielberg, Ph.D. for clinical assessments, RainerKniesche for technical assistance, and acknowledge accessto the MR scanner kindly provided by Charite Universitats-medizin Berlin.

Author Contributions

C.K., T.E. and N.K. designed the experiment, and C.K. col-lected the data. J.C.B. and C.K. analyzed the data, D.D.D.,C.L., T.D., R.G. and N.K. contributed to data interpretationand discussion, and J.C.B. and C.K. wrote the paper. All ofthe authors contributed to the preparation of the manuscript.

Author Disclosure Statement

The authors declare no competing financial interests.

References

Alexander GE, DeLong MR, Strick PL. 1986. Parallel organizationof functionally segregated circuits linking basal ganglia andcortex. Annu Rev Neurosci 9:357–381.

200 BEUCKE ET AL.

Alonso P, Menchon JM, Jimenez S, Segalas J, Mataix-Cols D, Jaur-rieta N, Labad J, Vallejo J, Cardoner N, Pujol J. 2008. Person-ality dimensions in obsessive-compulsive disorder: relation toclinical variables. Psychiatry Res 15:159–168.

Banks SJ, Eddy KT, Angstadt M, Nathan PJ, Phan KL. 2007.Amygdala-frontal connectivity during emotion regulation.Soc Cogn Affect Neurosci 2:303–312.

Baxter LR, Jr., Phelps ME, Mazziotta JC, Guze BH, Schwartz JM,Selin CE. 1987. Local cerebral glucose metabolic rates in obses-sive-compulsive disorder. A comparison with rates in unipo-lar depression and in normal controls. Arch Gen Psychiatry44:211–218.

Beck AT, Ward CH, Mendelson M, Mock J, Erbaugh J. 1961. Aninventory for measuring depression. Arch Gen Psychiatry4:561–571.

Breiter HC, Rauch SL, Kwong KK, Baker JR, Weisskoff RM, Ken-nedy DN, Kendrick AD, Davis TL, Jiang A, Cohen MS, SternCE, Belliveau JW, Baer L, O’Sullivan RL, Savage CR, JenikeMA, Rosen BR. 1996. Functional magnetic resonance imagingof symptom provocation in obsessive-compulsive disorder.Arch Gen Psychiatry 53:595–606.

Chamberlain SR, Menzies L, Hampshire A, Suckling J, FinebergNA, del Campo N, Aitken M, Craig K, Owen AM, BullmoreET, Robbins TW, Sahakian BJ. 2008. Orbitofrontal dysfunctionin patients with obsessive-compulsive disorder and their un-affected relatives. Science 321:421–422.

Das P, Kemp AH, Liddell BJ, Brown KJ, Olivieri G, Peduto A,Gordon E, Williams LM. 2005. Pathways for fear perception:modulation of amygdala activity by thalamo-cortical systems.Neuroimage 26:141–148.

Debener S, Ullsperger M, Siegel M, Fiehler K, von Cramon DY,Engel AK. 2005. Trial-by-trial coupling of concurrent electro-encephalogram and functional magnetic resonance imagingidentifies the dynamics of performance monitoring. J Neuro-sci 25:11730–11737.

Del Casale A, Kotzalidis GD, Rapinesi C, Serata D, Ambrosi E,Simonetti A, Pompili M, Ferracuti S, Tatarelli R, Girardi P.2011. Functional neuroimaging in obsessive-compulsive dis-order. Neuropsychobiology 64:61–85.

Denys D, Mantione M, Figee M, van den Munckhof P, Koersel-man F, Westenberg H, Bosch A, Schuurman R. 2010. Deepbrain stimulation of the nucleus accumbens for treatment-refractory obsessive-compulsive disorder. Arch Gen Psychia-try 67:1061–1068.

Dougherty DD, Baer L, Cosgrove GR, Cassem EH, Price BH,Nierenberg AA, Jenike MA, Rauch SL. 2002. Prospectivelong-term follow-up of 44 patients who received cingulotomyfor treatment-refractory obsessive-compulsive disorder. AmJ Psychiatry 159:269–275.

Endrass T, Klawohn J, Schuster F, Kathmann N. 2008. Overactiveperformance monitoring in obsessive-compulsive disorder:ERP evidence from correct and erroneous reactions. Neuro-psychologia 46:1877–1887.

Endrass T, Schuermann B, Kaufmann C, Spielberg R, Kniesche R,Kathmann N. 2010. Performance monitoring and error signif-icance in patients with obsessive-compulsive disorder. BiolPsychol 84:257–263.

First MB, Spitzer RL, Gibbon M, Williams JBW. 1996. StructuredClinical Interview for DSM-IV Axis 1 Disorders. Washington,DC: American Psychiatric Press.

Fitzgerald KD, Stern ER, Angstadt M, Nicholson-Muth KC, May-nor MR, Welsh RC, Hanna GL, Taylor SF. 2010. Altered func-tion and connectivity of the medial frontal cortex in pediatricobsessive-compulsive disorder. Biol Psychiatry 68:1039–1047.

Foa EB, Huppert JD, Leiberg S, Langner R, Kichic R, Hajcak G,Salkovskis PM. 2002. The Obsessive-Compulsive Inventory:development and validation of a short version. PsycholAssess 14:485–496.

Friston KJ. 2011. Functional and effective connectivity: a review.Brain Connect 1:13–36.

Friston KJ, Buechel C, Fink GR, Morris J, Rolls E, Dolan RJ. 1997.Psychophysiological and modulatory interactions in neuroi-maging. Neuroimage 6:218–229.

Friston KJ, Harrison L, Penny W. 2003. Dynamic causal model-ing. Neuroimage 19:1273–1302.

Gitelman DR, Penny WD, Ashburner J, Friston KJ. 2003. Model-ing regional and psychophysiologic interactions in fMRI: theimportance of hemodynamic deconvolution. Neuroimage19:200–207.

Goodman WK, Price LH, Rasmussen SA, Mazure C, FleischmannRL, Hill CL, Heninger GR, Charney DS. 1989. The Yale-BrownObsessive Compulsive Scale. I. Development, use, and reli-ability. Arch Gen Psychiatry 46:1006–1011.

Graybiel AM, Rauch SL. 2000. Toward a neurobiology of obses-sive-compulsive disorder. Neuron 28:343–347.

Guyer AE, Lau JY, McClure-Tone EB, Parrish J, Shiffrin ND, Rey-nolds RC, Chen G, Blair RJ, Leibenluft E, Fox NA, Ernst M,Pine DS, Nelson EE. 2008. Amygdala and ventrolateral pre-frontal cortex function during anticipated peer evaluation inpediatric social anxiety. Arch Gen Psychiatry 65:1303–1312.

Haber SN. 2003. The primate basal ganglia: parallel and integra-tive networks. J Chem Neuroanat 26:317–330.

Harrison BJ, Soriano-Mas C, Pujol J, Ortiz H, Lopez-Sola M,Hernandez-Ribas R, Deus J, Alonso P, Yucel M, Pantelis C,Menchon JM, Cardoner N. 2009. Altered corticostriatal func-tional connectivity in obsessive-compulsive disorder. ArchGen Psychiatry 66:1189–1200.

Kasahara M, Menon DK, Salmond CH, Outtrim JG, TaylorTavares JV, Carpenter TA, Pickard JD, Sahakian BJ, Stamata-kis EA. 2010. Altered functional connectivity in the motor net-work after traumatic brain injury. Neurology 75:168–176.

Knutson B, Adams CM, Fong GW, Hommer D. 2001. Anticipa-tion of increasing monetary reward selectively recruits nu-cleus accumbens. J Neurosci 21:RC159.

Knutson B, Westdorp A, Kaiser E, Hommer D. 2000. fMRI visu-alization of brain activity during a monetary incentive delaytask. Neuroimage 12:20–27.

Kopell BH, Greenberg BD. 2008. Anatomy and physiology of thebasal ganglia: implications for DBS in psychiatry. NeurosciBiobehav Rev 32:408–422.

Kosslyn SM. 1999. If neuroimaging is the answer, what is thequestion? Philos Trans R Soc Lond B Biol Sci 354:1283–1294.

Kunishio K, Haber SN. 1994. Primate cingulostriatal projection:limbic striatal versus sensorimotor striatal input. J Comp Neu-rol 350:337–356.

Lanius RA, Williamson PC, Densmore M, Boksman K, NeufeldRW, Gati JS, Menon RS. 2004. The nature of traumatic mem-ories: a 4-T FMRI functional connectivity analysis. Am J Psy-chiatry 161:36–44.

Linnman C, Beucke JC, Jensen KB, Gollub RL, Kong J. 2012. Sexsimilarities and differences in pain-related periaqueductalgray connectivity. Pain 153:444–454.

Magno E, Foxe JJ, Molholm S, Robertson IH, Garavan H. 2006.The anterior cingulate and error avoidance. J Neurosci26:4769–4773.

Maia TV, Frank MJ. 2011. From reinforcement learning modelsto psychiatric and neurological disorders. Nat Neurosci 14:154–162.

ALTERED CINGULOSTRIATAL COUPLING IN OCD 201

Menzies L, Chamberlain SR, Laird AR, Thelen SM, Sahakian BJ,Bullmore ET. 2008a. Integrating evidence from neuroimagingand neuropsychological studies of obsessive-compulsivedisorder: the orbitofronto-striatal model revisited. NeurosciBiobehav Rev 32:525–549.

Menzies L, Williams GB, Chamberlain SR, Ooi C, Fineberg N,Suckling J, Sahakian BJ, Robbins TW, Bullmore ET. 2008b.White matter abnormalities in patients with obsessive-compulsive disorder and their first-degree relatives. AmJ Psychiatry 165:1308–1315.

Mobbs D, Marchant JL, Hassabis D, Seymour B, Tan G, Gray M,Petrovic P, Dolan RJ, Frith CD. 2009. From threat to fear: theneural organization of defensive fear systems in humans.J Neurosci 29:12236–12243.

Oldfield RC. 1971. The assessment and analysis of handedness:the Edinburgh inventory. Neuropsychologia 9:97–113.

Pujol J, Soriano-Mas C, Alonso P, Cardoner N, Menchon JM,Deus J, Vallejo J. 2004. Mapping structural brain alterationsin obsessive-compulsive disorder. Arch Gen Psychiatry 61:720–730.

Rauch SL, Dougherty DD, Malone D, Rezai A, Friehs G, Fisch-man AJ, Alpert NM, Haber SN, Stypulkowski PH, Rise MT,Rasmussen SA, Greenberg BD. 2006. A functional neuroimag-ing investigation of deep brain stimulation in patients withobsessive-compulsive disorder. J Neurosurg 104:558–565.

Rauch SL, Jenike MA, Alpert NM, Baer L, Breiter HC, Savage CR,Fischman AJ. 1994. Regional cerebral blood flow measuredduring symptom provocation in obsessive-compulsive disor-der using oxygen 15-labeled carbon dioxide and positronemission tomography. Arch Gen Psychiatry 51:62–70.

Rauch SL, Kim H, Makris N, Cosgrove GR, Cassem EH, SavageCR, Price BH, Nierenberg AA, Shera D, Baer L, Buchbinder B,Caviness VS, Jr., Jenike MA, Kennedy DN. 2000. Volume re-duction in the caudate nucleus following stereotactic place-ment of lesions in the anterior cingulate cortex in humans: amorphometric magnetic resonance imaging study. J Neuro-surg 93:1019–1025.

Remijnse PL, Nielen MM, van Balkom AJ, Cath DC, van Oppen P,Uylings HB, Veltman DJ. 2006. Reduced orbitofrontal-striatal activity on a reversal learning task in obsessive-compulsive disorder. Arch Gen Psychiatry 63:1225–1236.

Riesel A, Endrass T, Kaufmann C, Kathmann N. 2011. Overactiveerror-related brain activity as a candidate endophenotype forobsessive-compulsive disorder: evidence from unaffectedfirst-degree relatives. Am J Psychiatry 168:317–324.

Rotge JY, Guehl D, Dilharreguy B, Tignol J, Bioulac B, Allard M,Burbaud P, Aouizerate B. 2009. Meta-analysis of brain volumechanges in obsessive-compulsive disorder. Biol Psychiatry65:75–83.

Ruck C, Karlsson A, Steele JD, Edman G, Meyerson BA, EricsonK, Nyman H, Asberg M, Svanborg P. 2008. Capsulotomy forobsessive-compulsive disorder: long-term follow-up of 25 pa-tients. Arch Gen Psychiatry 65:914–921.

Rushworth MF, Behrens TE. 2008. Choice, uncertainty and valuein prefrontal and cingulate cortex. Nat Neurosci 11:389–397.

Salkovskis PM. 1985. Obsessional-compulsive problems: a cogni-tive-behavioural analysis. Behav Res Ther 23:571–583.

Saxena S, Brody AL, Schwartz JM, Baxter LR. 1998. Neuroimag-ing and frontal-subcortical circuitry in obsessive-compulsivedisorder. Br J Psychiatry Suppl 35:26–37.

Schmidt KH, Metzler P. 1992. Wortschatztest (WST). Weinheim:Beltz.

Shackman AJ, Salomons TV, Slagter HA, Fox AS, Winter JJ,Davidson RJ. 2011. The integration of negative affect, painand cognitive control in the cingulate cortex. Nat Rev Neuro-sci 12:154–167.

Simon D, Kaufmann C, Musch K, Kischkel E, Kathmann N. 2010.Fronto-striato-limbic hyperactivation in obsessive-compulsivedisorder during individually tailored symptom provocation.Psychophysiology 47:728–738.

Spielberger CD, Gorsuch RL, Lushene RE. 1970. Manual for theState-Trait Anxiety Inventory. Palo Alto: Consulting Psycholo-gists Press.

Stephan KE. 2004. On the role of general system theory for func-tional neuroimaging. J Anat 205:443–470.

Stephan KE, Penny WD, Moran RJ, den Ouden HE, Daunizeau J,Friston KJ. 2010. Ten simple rules for dynamic causal model-ing. Neuroimage 49:3099–3109.

Swedo SE, Schapiro MB, Grady CL, Cheslow DL, Leonard HL,Kumar A, Friedland R, Rapoport SI, Rapoport JL. 1989.Cerebral glucose metabolism in childhood-onset obsessive-compulsive disorder. Arch Gen Psychiatry 46:518–523.

Szeszko PR, Ardekani BA, Ashtari M, Malhotra AK, RobinsonDG, Bilder RM, Lim KO. 2005. White matter abnormalitiesin obsessive-compulsive disorder: a diffusion tensor imagingstudy. Arch Gen Psychiatry 62:782–790.

Vogt BA. 2005. Pain and emotion interactions in subregions of thecingulate gyrus. Nat Rev Neurosci 6:533–544.

Vogt BA, Pandya DN. 1987. Cingulate cortex of the rhesus mon-key: II. Cortical afferents. J Comp Neurol 262:271–289.

Wang L, Metzak PD, Woodward TS. 2011. Aberrant connectivityduring self-other source monitoring in schizophrenia. Schiz-ophr Res 125:136–142.

Whiteside SP, Port JD, Abramowitz JS. 2004. A meta-analysis offunctional neuroimaging in obsessive-compulsive disorder.Psychiatry Res 132:69–79.

Williams LM, Das P, Liddell BJ, Kemp AH, Rennie CJ, Gordon E.2006. Mode of functional connectivity in amygdala pathwaysdissociates level of awareness for signals of fear. J Neurosci26:9264–9271.

Address correspondence to:Christian Kaufmann

Humboldt-Universitat zu BerlinDepartment of Psychology

Rudower Chaussee 18D-12489 Berlin

Germany

E-mail: christian.kaufmann@psychologie.hu-berlin.de

202 BEUCKE ET AL.